Identification of prognostic signature of non–small cell lung cancer based on TCGA methylation data

Abstract

AbstractNon–small lung cancer (NSCLC) is a common malignant disease with very poor outcome. Accurate prediction of prognosis can better guide patient risk stratification and treatment decision making, and could optimize the outcome. Utilizing clinical and methylation/expression data in The Cancer Genome Atlas (TCGA), we conducted comprehensive evaluation of early-stage NSCLC to identify a methylation signature for survival prediction. 349 qualified cases of NSCLC with curative surgery were included and further grouped into the training and validation cohorts. We identified 4000 methylation loci with prognostic influence on univariate and multivariate regression analysis in the training cohort. KEGG pathway analysis was conducted to identify the key pathway. Hierarchical clustering and WGCNA co-expression analysis was performed to classify the sample phenotype and molecular subtypes. Hub 5′-C-phosphate-G-3′ (CpG) loci were identified by network analysis and then further applied for the construction of the prognostic signature. The predictive power of the prognostic model was further validated in the validation cohort. Based on clustering analysis, we identified 6 clinical molecular subtypes, which were associated with different clinical characteristics and overall survival; clusters 4 and 6 demonstrated the best and worst outcomes. We identified 17 hub CpG loci, and their weighted combination was used for the establishment of a prognostic model (RiskScore). The RiskScore significantly correlated with post-surgical outcome; patients with a higher RiskScore have worse overall survival in both the training and validation cohorts (P < 0.01). We developed a novel methylation signature that can reliably predict prognosis for patients with NSCLC.